pith. sign in

arxiv: 2605.17039 · v1 · pith:UJBTDMCGnew · submitted 2026-05-16 · 💻 cs.LG · cs.CE

Privacy-Preserving Generation Fraud Detection for Distributed Photovoltaic Systems: A Solar Irradiance-Fused Federated Learning Framework

Xiaolu Chen , Chenghao Huang , Yanru Zhang , Hao Wang This is my paper

Pith reviewed 2026-05-19 20:13 UTC · model grok-4.3

classification 💻 cs.LG cs.CE
keywords federated learningphotovoltaic fraud detectionsolar irradiance fusionco-attention mechanismprivacy preservationclass imbalance handlingdistributed energy systems
0
0 comments X

The pith

A federated learning framework fuses solar irradiance data via co-attention to detect photovoltaic generation fraud while keeping household data private across communities.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a distributed framework for detecting fraud in photovoltaic generation reports from residential systems. It uses federated learning so that each community trains a local model on its own data and shares only model updates, avoiding central collection of sensitive household information. Each local detector fuses PV output readings with weather and irradiance data through a co-attention mechanism to identify suspicious discrepancies. Prototype alignment across communities helps correct for the rarity of fraud examples. Experiments on real residential PV data show the approach beats existing federated methods while scaling to different community sizes and remaining robust when fraud cases are scarce.

Core claim

The central claim is that integrating a co-attention mechanism to fuse PV generation and solar irradiance data within a federated learning setup, augmented by prototype alignment to address class imbalance, allows effective and privacy-preserving detection of PV generation fraud across multiple household communities, outperforming state-of-the-art federated learning methods on real-world datasets.

What carries the argument

co-attention mechanism that fuses PV generation data with solar irradiance and weather information, combined with prototype alignment in federated aggregation

If this is right

  • Local models can detect fraud by spotting mismatches between reported generation and actual solar conditions without centralizing raw data.
  • Cross-community collaboration through parameter and prototype sharing improves detection while keeping household privacy intact.
  • The method scales effectively as the number of communities grows.
  • Prototype alignment provides robustness when fraudulent samples are much rarer than normal ones.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar fusion and alignment techniques could extend to fraud detection in other intermittent renewable sources like wind turbines.
  • If the discrepancy detection proves reliable, regulators might adopt irradiance-based audits as a standard for verifying PV incentives.
  • Testing on synthetic fraud injections of varying sophistication would clarify the limits of the co-attention approach.

Load-bearing premise

Discrepancies between reported PV generation and fused solar irradiance or weather data can reliably signal fraud, and aligning prototypes across communities corrects class imbalance without altering genuine generation patterns.

What would settle it

Running the framework on a held-out residential PV dataset where known fraud cases are labeled by independent inspection and checking whether detection precision drops below that of centralized or non-fused baselines.

Figures

Figures reproduced from arXiv: 2605.17039 by Chenghao Huang, Hao Wang, Xiaolu Chen, Yanru Zhang.

Figure 1
Figure 1. Figure 1: In the i-th community, for each day d in a total of D days, a daily horizon is divided into T discrete time slots. We denote the actual generation of a solar prosumer at time slot t as x APVG i,d,t and the generation reported to the utility company via a smart meter as x PVG i,d,t. Under normal conditions, these two values coincide, i.e., x PVG i,d,t = x APVG i,d,t . However, in cases of PV generation frau… view at source ↗
Figure 1
Figure 1. Figure 1: The FL paradigm for distributed PVG-FD. daily reported-generation time series as x PVG i,d = {x PVG i,d,t} T t=1. To indicate whether fraud occurs on a given day d in the i-th community, we introduce a binary variable yi,d defined by yi,d = ( 1, if fraud occurs on day d, 0, otherwise. (2) Furthermore, as the characteristics of individual PV panels may remain unknown, it is beneficial to incorporate weather… view at source ↗
Figure 2
Figure 2. Figure 2: The architecture of the i-th community’s PVG-FD model carrying out detection on the d-th day of the j-th prosumer. 1) Data Pre-Processing: At the local end, each community performs PVG-FD using recent time-series data from smart meter readings. Due to privacy constraints, detailed PV system specifications are not accessible. To compensate for this lack of system-level information, external weather data, in… view at source ↗
Figure 3
Figure 3. Figure 3: Overview of the FL protocol in the proposed PVG-FD framework. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of normal and malicious solar generation data for [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Learning curves of local models under the FedAvg framework across [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Learning curves of FL frameworks across four evaluation metrics. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Performance comparison under varying community counts with a [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Performance of FL frameworks under varying PV generation fraud [PITH_FULL_IMAGE:figures/full_fig_p012_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Daily fraud probabilities for prosumer ID 194 over one year, with ground truth and predictions from different federated configurations. [PITH_FULL_IMAGE:figures/full_fig_p013_9.png] view at source ↗
read the original abstract

The wide adoption of residential photovoltaic (PV) systems introduces new challenges for generation fraud detection (FD). Unlike traditional electricity theft detection, which focuses on electricity consumption-side behavior, PV generation fraud detection (PVG-FD) is complicated by the inherent intermittency and uncertainty of PV generation. The distributed nature of PV systems poses further challenges for centralized PVG-FD approaches due to scalability and privacy concerns. This paper develops a privacy-preserving distributed PVG-FD framework based on federated learning (FL). In this framework, a utility company manages multiple household communities, where each of which is equipped with a local detector. The framework integrates a novel detection model architecture with privacy-preserving global collaboration. Each community's local model fuses PV generation and weather data via a co-attention mechanism to detect discrepancies critical for PVG-FD. The FL framework enables cross-community collaboration by aggregating model parameters and prototypes, leveraging global knowledge sharing with local refinement while preserving privacy. It also uses prototype alignment to address class imbalance by enhancing fraud sample representation. Extensive experiments on a real-world residential PV dataset validate the effectiveness of the developed method and demonstrate that it outperforms state-of-the-art FL methods across various scenarios. The results also show its scalability across varying community sizes and strong robustness to class imbalance.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper proposes a privacy-preserving federated learning framework for generation fraud detection in distributed residential photovoltaic systems. Local detectors at each community fuse PV generation data with solar irradiance and weather information using a co-attention mechanism to identify discrepancies as fraud signals. The FL component aggregates model parameters and prototypes across communities for global knowledge sharing while preserving privacy, with prototype alignment used to mitigate class imbalance. Experiments on a real-world residential PV dataset are reported to show outperformance over state-of-the-art FL methods, scalability across community sizes, and robustness to class imbalance.

Significance. If the empirical claims hold after addressing validation details, the work would offer a practical contribution at the intersection of federated learning and renewable energy security. The co-attention fusion and prototype alignment address domain-specific challenges of intermittency and imbalance in a distributed, privacy-sensitive setting, potentially informing utility-scale deployments for PV fraud detection.

major comments (2)
  1. [Methods (local model architecture)] Local detector / co-attention description: The framework treats discrepancies between reported PV generation and fused irradiance/weather data as reliable fraud indicators, yet provides no explicit controls, sensitivity analysis, or discussion of confounders such as local shading, panel degradation, forecast error, or sensor noise. This assumption is load-bearing for the central claim, as the reported outperformance and robustness on real-world data would be undermined if such factors produce false positives not isolated by the co-attention or alignment steps.
  2. [Experiments section] Experiments and results: The abstract and claims assert quantitative outperformance, scalability, and robustness to class imbalance on a real-world dataset, but the provided description lacks specific metrics (e.g., precision, recall, F1, or AUC values), baseline details, ablation results, or information on how fraud labels were obtained and validated. Without these, the support for the effectiveness claims remains high-level and difficult to verify.
minor comments (1)
  1. [Abstract] The acronym PVG-FD is used in the abstract without prior expansion; spell out 'PV generation fraud detection' on first use for clarity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback on our manuscript. We address each major comment below with honest responses and indicate where revisions will be made to strengthen the work.

read point-by-point responses

  1. Referee: [Methods (local model architecture)] Local detector / co-attention description: The framework treats discrepancies between reported PV generation and fused irradiance/weather data as reliable fraud indicators, yet provides no explicit controls, sensitivity analysis, or discussion of confounders such as local shading, panel degradation, forecast error, or sensor noise. This assumption is load-bearing for the central claim, as the reported outperformance and robustness on real-world data would be undermined if such factors produce false positives not isolated by the co-attention or alignment steps.

    Authors: We acknowledge that environmental and operational factors such as local shading, panel degradation, forecast error, and sensor noise can produce discrepancies that might mimic fraud signals. The co-attention mechanism is intended to learn joint representations that help the model prioritize fraud-related patterns over natural intermittency, but we agree that the current manuscript lacks an explicit discussion or sensitivity analysis of these confounders. In the revision, we will add a dedicated subsection on potential confounders and include new sensitivity experiments (e.g., simulated shading and noise injection) to evaluate false-positive rates and demonstrate that performance remains stable. revision: yes

  2. Referee: [Experiments section] Experiments and results: The abstract and claims assert quantitative outperformance, scalability, and robustness to class imbalance on a real-world dataset, but the provided description lacks specific metrics (e.g., precision, recall, F1, or AUC values), baseline details, ablation results, or information on how fraud labels were obtained and validated. Without these, the support for the effectiveness claims remains high-level and difficult to verify.

    Authors: The full manuscript contains quantitative results with specific metrics (precision, recall, F1, AUC) and comparisons against baselines such as FedAvg and other FL methods, along with ablation studies on the co-attention and prototype alignment modules. Fraud labels originate from verified utility-reported cases in the real-world residential PV dataset. However, we accept that these elements could be presented more prominently and with greater clarity. We will revise the abstract, add a summary table of key metrics, and expand the dataset description to include validation details for the labels. revision: partial

Circularity Check

0 steps flagged

No circularity: empirical engineering framework with external validation

full rationale

The paper presents a federated learning framework for PV generation fraud detection that fuses solar irradiance/weather data via co-attention and uses prototype alignment to handle imbalance. Central claims rest on experimental results from a real-world residential PV dataset showing outperformance over SOTA FL methods, scalability, and robustness. No mathematical derivations, predictions, or first-principles results appear that reduce to inputs by construction. The contribution is self-contained as an applied ML engineering effort with independent external data validation; no self-definitional steps, fitted inputs renamed as predictions, or load-bearing self-citations are present in the derivation chain.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework rests on standard federated learning assumptions about non-IID data across communities and the premise that weather data is available and aligned with generation readings; no new free parameters or invented entities are introduced in the abstract.

axioms (2)
  • domain assumption Local PV generation data can be paired with publicly available or locally measured solar irradiance and weather data to reveal fraud-indicating discrepancies.
    Invoked in the description of the co-attention fusion mechanism.
  • domain assumption Federated averaging of model parameters and prototypes preserves privacy while enabling useful global knowledge transfer.
    Core premise of the FL component.

pith-pipeline@v0.9.0 · 5770 in / 1388 out tokens · 33413 ms · 2026-05-19T20:13:04.550723+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

44 extracted references · 44 canonical work pages

  1. [1]

    Renewables 2024,

    International Energy Agency, “Renewables 2024,” 2023. [Online]. Available: https://www.iea.org/reports/renewables-2024

    work page 2024

  2. [2]

    Renewable energy market update,

    ——, “Renewable energy market update,” 2023. [Online]. Available: https://www.iea.org/reports/renewable-energy-market-update-june-2023

    work page 2023

  3. [3]

    A review of cyber security risks of power systems: from static to dynamic false data attacks,

    Y . Xu, “A review of cyber security risks of power systems: from static to dynamic false data attacks,”Protection and Control of Modern Power Systems, vol. 5, no. 1, p. 19, 2020

    work page 2020

  4. [4]

    Detection methods in smart meters for electricity thefts: A survey,

    X. Xia, Y . Xiao, W. Liang, and J. Cui, “Detection methods in smart meters for electricity thefts: A survey,”Proceedings of the IEEE, vol. 110, no. 2, pp. 273–319, 2022

    work page 2022

  5. [5]

    Deep learning detection of electricity theft cyber-attacks in renewable distributed generation,

    M. Ismail, M. F. Shaaban, M. Naidu, and E. Serpedin, “Deep learning detection of electricity theft cyber-attacks in renewable distributed generation,”IEEE Transactions on Smart Grid, vol. 11, no. 4, pp. 3428– 3437, 2020

    work page 2020

  6. [6]

    Evaluating detectors on optimal attack vectors that enable electricity theft and der fraud,

    V . B. Krishna, C. A. Gunter, and W. H. Sanders, “Evaluating detectors on optimal attack vectors that enable electricity theft and der fraud,” IEEE Journal of Selected Topics in Signal Processing, vol. 12, no. 4, pp. 790–805, 2018

    work page 2018

  7. [7]

    Energy theft detection in smart grids: Taxonomy, compar- ative analysis, challenges, and future research directions,

    M. Ahmed, A. Khan, M. Ahmed, M. Tahir, G. Jeon, G. Fortino, and F. Piccialli, “Energy theft detection in smart grids: Taxonomy, compar- ative analysis, challenges, and future research directions,”IEEE/CAA Journal of Automatica Sinica, vol. 9, no. 4, pp. 578–600, 2022

    work page 2022

  8. [8]

    Classification- based electricity theft detection on households with photovoltaic gen- eration and net metering,

    R. M. P. Maala, A. M. B. Rebamba, and A. E. D. Tio, “Classification- based electricity theft detection on households with photovoltaic gen- eration and net metering,” inTENCON 2023 - 2023 IEEE Region 10 Conference (TENCON), 2023, pp. 1094–1099

    work page 2023

  9. [9]

    Fine-tuned rnn-based detector for electricity theft attacks in smart grid generation domain,

    M. E. Eddin, A. Albaseer, M. Abdallah, S. Bayhan, M. K. Qaraqe, S. Al- Kuwari, and H. Abu-Rub, “Fine-tuned rnn-based detector for electricity theft attacks in smart grid generation domain,”IEEE Open Journal of the Industrial Electronics Society, vol. 3, pp. 733–750, 2022

    work page 2022

  10. [10]

    A novel multi-classification method for photovoltaic electricity theft behavior with low false detection rate,

    C. Chen, R. Xia, and H. Hu, “A novel multi-classification method for photovoltaic electricity theft behavior with low false detection rate,” Measurement, vol. 239, p. 115461, 2025

    work page 2025

  11. [11]

    A data augmentation method for distributed photovoltaic electricity theft using wasserstein generative adversarial network,

    J. Li, W. Liao, R. Yang, and Z. Chen, “A data augmentation method for distributed photovoltaic electricity theft using wasserstein generative adversarial network,” in2021 IEEE 5th Conference on Energy Internet and Energy System Integration (EI2), 2021, pp. 3132–3137

    work page 2021

  12. [12]

    Data-driven detection of electricity theft cyberattacks in pv generation,

    M. Shaaban, U. Tariq, M. Ismail, N. A. Almadani, and M. Mokhtar, “Data-driven detection of electricity theft cyberattacks in pv generation,” IEEE Systems Journal, vol. 16, no. 2, pp. 3349–3359, 2022

    work page 2022

  13. [13]

    Vulnerability analysis of demand- response with renewable energy integration in smart grids to cyber attacks and online detection methods,

    D. Tang, Y .-P. Fang, and E. Zio, “Vulnerability analysis of demand- response with renewable energy integration in smart grids to cyber attacks and online detection methods,”Reliability Engineering & System Safety, vol. 235, p. 109212, 2023

    work page 2023

  14. [14]

    Smart energy guardian: A hybrid deep learning model for detecting fraudulent pv generation,

    X. Chen, C. Huang, Y . Zhang, and H. Wang, “Smart energy guardian: A hybrid deep learning model for detecting fraudulent pv generation,” in2024 IEEE International Smart Cities Conference (ISC2), 2024, pp. 1–6

    work page 2024

  15. [15]

    Communication-Efficient Learning of Deep Networks from Decen- tralized Data,

    B. McMahan, E. Moore, D. Ramage, S. Hampson, and B. A. y. Arcas, “Communication-Efficient Learning of Deep Networks from Decen- tralized Data,” inProceedings of the 20th International Conference on Artificial Intelligence and Statistics, ser. Proceedings of Machine Learning Research, vol. 54. PMLR, 20–22 Apr 2017, pp. 1273–1282

    work page 2017

  16. [16]

    Decentralized privacy- preserving electricity theft detection for distribution system operators,

    X. Wang, H. Xie, L. Tang, C. Chen, and Z. Bie, “Decentralized privacy- preserving electricity theft detection for distribution system operators,” IEEE Transactions on Smart Grid, vol. 15, no. 2, pp. 2179–2190, 2024

    work page 2024

  17. [17]

    Electricity theft detection in the presence of prosumers using a cluster-based multi-feature detection model,

    A. Alromih, J. A. Clark, and P. Gope, “Electricity theft detection in the presence of prosumers using a cluster-based multi-feature detection model,” in2021 IEEE International Conference on Communications, Control, and Computing Technologies for Smart Grids (SmartGrid- Comm), 2021, pp. 339–345

    work page 2021

  18. [18]

    Data-driven energy theft detection in modern power grids,

    A. Althobaiti, A. Jindal, and A. K. Marnerides, “Data-driven energy theft detection in modern power grids,” inProceedings of the Twelfth ACM International Conference on Future Energy Systems, ser. e-Energy ’21. New York, NY , USA: Association for Computing Machinery, 2021, p. 39–48. [Online]. Available: https://doi.org/10.1145/3447555.3464852

    work page doi:10.1145/3447555.3464852 2021

  19. [19]

    Power theft detection in smart grids using quantum machine learning,

    K. Blazakis, N. Schetakis, M. M. Badr, D. Aghamalyan, K. Stavrakakis, and G. Stavrakakis, “Power theft detection in smart grids using quantum machine learning,”IEEE Access, vol. 13, pp. 61 511–61 525, 2025

    work page 2025

  20. [20]

    Adaptive energy theft detection in smart grids using self-learning with dual neural network,

    A. Althobaiti, C. Rotsos, and A. K. Marnerides, “Adaptive energy theft detection in smart grids using self-learning with dual neural network,” IEEE Transactions on Industrial Informatics, vol. 20, no. 2, pp. 2776– 2786, 2024

    work page 2024

  21. [21]

    Deep learning-based electricity theft detection for active distribution networks with high photovoltaic penetration,

    S. Liu, J. Lv, P. Lv, and Y . Cheng, “Deep learning-based electricity theft detection for active distribution networks with high photovoltaic penetration,”Computers and Electrical Engineering, vol. 126, p. 110551, 2025. [Online]. Available: https://www.sciencedirect.com/scie nce/article/pii/S004579062500494X

    work page 2025

  22. [22]

    Coordinated data falsification attack detection in the domain of distributed generation using deep learning,

    N. Bhusal, M. Gautam, R. M. Shukla, M. Benidris, and S. Sengupta, “Coordinated data falsification attack detection in the domain of distributed generation using deep learning,”International Journal of Electrical Power & Energy Systems, vol. 134, p. 107345, 2022. [Online]. Available: https://www.sciencedirect.com/science/article/pii/ S0142061521005846

    work page 2022

  23. [23]

    Data driven electricity theft detection based on federated learning,

    F. Han, X. Wang, Y . Chen, J. Qiao, L. Gao, and X. Xin, “Data driven electricity theft detection based on federated learning,” in2023 IEEE 7th Conference on Energy Internet and Energy System Integration (EI2), 2023, pp. 4865–4870

    work page 2023

  24. [24]

    An efficient se- quential decentralized federated progressive channel pruning strategy for 15 smart grid electricity theft detection,

    F. Guo, S. Li, H. Yang, C. Dong, Y . Chen, and G. Li, “An efficient se- quential decentralized federated progressive channel pruning strategy for 15 smart grid electricity theft detection,”IEEE Transactions on Industrial Informatics, vol. 21, no. 3, pp. 2393–2402, 2025

    work page 2025

  25. [25]

    Fuse: a federated learning and u-shape split learning-based electricity theft detection framework,

    X. Li, N. Wang, L. Zhu, S. Yuan, and Z. Guan, “Fuse: a federated learning and u-shape split learning-based electricity theft detection framework,”Science China Information Sciences, vol. 67, no. 4, p. 149302, 2024

    work page 2024

  26. [26]

    Privacy-preserving and efficient decentralized federated learning-based energy theft detector,

    M. I. Ibrahem, M. Mahmoud, M. M. Fouda, B. M. ElHalawany, and W. Alasmary, “Privacy-preserving and efficient decentralized federated learning-based energy theft detector,” inGLOBECOM 2022 - 2022 IEEE Global Communications Conference, 2022, pp. 287–292

    work page 2022

  27. [27]

    An energy theft detection framework with privacy protection for smart grid,

    R. Xie, “An energy theft detection framework with privacy protection for smart grid,” in2023 International Joint Conference on Neural Networks (IJCNN), 2023, pp. 1–7

    work page 2023

  28. [28]

    A privacy-preserving heterogeneous federated learning framework with class imbalance learning for electricity theft detection,

    H. Wen, X. Liu, B. Lei, M. Yang, X. Cheng, and Z. Chen, “A privacy-preserving heterogeneous federated learning framework with class imbalance learning for electricity theft detection,”Applied Energy, vol. 378, p. 124789, 2025

    work page 2025

  29. [29]

    Feddetect: A novel privacy-preserving federated learning framework for energy theft detection in smart grid,

    M. Wen, R. Xie, K. Lu, L. Wang, and K. Zhang, “Feddetect: A novel privacy-preserving federated learning framework for energy theft detection in smart grid,”IEEE Internet of Things Journal, vol. 9, no. 8, pp. 6069–6080, 2022

    work page 2022

  30. [30]

    Cyberattack detection for electricity theft in smart grids via stacking ensemble gru op- timization algorithm using federated learning framework,

    J. Wang, Y . Si, Y . Zhu, K. Zhang, S. Yin, and B. Liu, “Cyberattack detection for electricity theft in smart grids via stacking ensemble gru op- timization algorithm using federated learning framework,”International Journal of Electrical Power & Energy Systems, vol. 157, p. 109848, 2024

    work page 2024

  31. [31]

    Fedcosr: Personalized federated learning with contrastive shareable representations for label heterogeneity in non-iid data,

    C. Huang, X. Chen, Y . Zhang, and H. Wang, “Fedcosr: Personalized federated learning with contrastive shareable representations for label heterogeneity in non-iid data,”IEEE Transactions on Cybernetics, vol. 55, no. 12, pp. 5841–5854, 2025

    work page 2025

  32. [32]

    Balancefl: Addressing class imbalance in long-tail federated learning,

    X. Shuai, Y . Shen, S. Jiang, Z. Zhao, Z. Yan, and G. Xing, “Balancefl: Addressing class imbalance in long-tail federated learning,” in2022 21st ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN), 2022, pp. 271–284

    work page 2022

  33. [33]

    Tackling data heterogeneity in federated learning with class prototypes,

    Y . Dai, Z. Chen, J. Li, S. Heinecke, L. Sun, and R. Xu, “Tackling data heterogeneity in federated learning with class prototypes,”Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, no. 6, pp. 7314–7322, Jun. 2023

    work page 2023

  34. [34]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, “Attention is all you need,” inProceedings of the 31st International Conference on Neural Information Processing Systems, ser. NIPS’17. Red Hook, NY , USA: Curran Associates Inc., 2017, p. 6000–6010

    work page 2017

  35. [35]

    Stacked cross attention for image-text matching,

    K.-H. Lee, X. Chen, G. Hua, H. Hu, and X. He, “Stacked cross attention for image-text matching,” inProceedings of the European Conference on Computer Vision (ECCV), September 2018

    work page 2018

  36. [36]

    How transferable are features in deep neural networks?

    J. Yosinski, J. Clune, Y . Bengio, and H. Lipson, “How transferable are features in deep neural networks?” inAdvances in Neural Infor- mation Processing Systems, Z. Ghahramani, M. Welling, C. Cortes, N. Lawrence, and K. Weinberger, Eds., vol. 27. Curran Associates, Inc., 2014

    work page 2014

  37. [37]

    Prototypical networks for few- shot learning,

    J. Snell, K. Swersky, and R. Zemel, “Prototypical networks for few- shot learning,” inAdvances in Neural Information Processing Systems, I. Guyon, U. V . Luxburg, S. Bengio, H. Wallach, R. Fergus, S. Vish- wanathan, and R. Garnett, Eds., vol. 30. Curran Associates, Inc., 2017

    work page 2017

  38. [38]

    Residential load and rooftop pv generation: an australian distribution network dataset,

    E. L. Ratnam, S. R. Weller, C. M. Kellett, and A. T. Murray, “Residential load and rooftop pv generation: an australian distribution network dataset,”International Journal of Sustainable Energy, vol. 36, no. 8, pp. 787–806, 2017

    work page 2017

  39. [39]

    Nsrdb: National solar radiation database

    National Renewable Energy Laboratory, “Nsrdb: National solar radiation database.” [Online]. Available: https://nsrdb.nrel.gov/

    work page

  40. [40]

    Electricity theft detection in ami using customers’ consumption patterns,

    P. Jokar, N. Arianpoo, and V . C. M. Leung, “Electricity theft detection in ami using customers’ consumption patterns,”IEEE Transactions on Smart Grid, vol. 7, no. 1, pp. 216–226, 2016

    work page 2016

  41. [41]

    Long short-term memory,

    S. Hochreiter, “Long short-term memory,”Neural Computation MIT- Press, 1997

    work page 1997

  42. [42]

    Long-term recurrent convolutional networks for visual recognition and description,

    J. Donahue, L. Anne Hendricks, S. Guadarrama, M. Rohrbach, S. Venu- gopalan, K. Saenko, and T. Darrell, “Long-term recurrent convolutional networks for visual recognition and description,” inProceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2015

    work page 2015

  43. [43]

    Reformer: The efficient trans- former,

    N. Kitaev, L. Kaiser, and A. Levskaya, “Reformer: The efficient trans- former,” inInternational Conference on Learning Representations, 2020

    work page 2020

  44. [44]

    Are transformers effective for time series forecasting?

    A. Zeng, M. Chen, L. Zhang, and Q. Xu, “Are transformers effective for time series forecasting?”Proceedings of the AAAI Conference on Artificial Intelligence, vol. 37, no. 9, pp. 11 121–11 128, Jun. 2023. Xiaolu Chenreceived the B.E. degree in Cyberspace Security and the M.S. degree in Computer Tech- nology from the School of Computer Science and Engineer...

    work page 2023